052C/052D Class Destroyers

bebops said:

055 is great against warship vs warship. but

In the modern warfare, they need a warship that can deal with hundreds of drones and UUV at once. Perhaps a warship with multiple laser system and CIWS.

A single warship that can destroy swarm satellite, water drones and air drones..

Click to expand...

Under what circumstances do you envision a blue water AAD Destroyer having to deal with "hundreds of drones and UUV at once"? And how big are these drones to begin with?

Cos the issues that pop immediately ...

Size -
If these drones are big then who in the world has the capability to sortie that many defacto cruise missiles at a fleet? And I use fleet because what are the odds that you have an 055 all by itself?
If these drones are small, then range is short, then how is that small drone gonna have the range to get to a blue water platform like the 055?

Speed -
Typical characteristic of a drone is they are slow. We're talking 100+kmph being typical. Contrast with traditional sub-sonic AShM threat that vector in at almost an order of magnitude faster.
Speed = engagement time. 10 times slower means each defending platform can engage up to 10 times that number of incoming drones vs subsonic AShM (hence the pucker factor behind high supersonic/hypersonics AShM)

You combine the above 2 (Size + Speed) ..
- If your drone is big+fast=Long Range, you have effectively a cruise missile. Again, who in the world has that magazine depth and capability to alpha strike 500+ cruise missiles at a CSG/SAG/ESG? Answer, probably only PLAN actually (and only within 1IC).
- If your drone is big+slow=Long Range, you have even more time for the target fleet to literally pick and choose how to defend against you. Refer to the recent Iranian strike against Israel - every single drone was picked off in their multi-hour transit to the target.
- If your drone is small+anything=Short Range, you can't get at a 055 that is OTH. So where are you launching from? Far easier for the 055 to sink your launch platform than deal with your "hundreds of drone".

You could of-course redesign a harpoon analog that instead of a unitary warhead, carries 10 switchblade like drones up to 40km from target fleet and then dispense them. But again, the 055 can simply engage your carrier AShM at 40+km range and be done with it using 2 AAMs.

Not to mention that a switchblade analog is still a slow drone which will take about 20 minutes to cover 40km. An eternity for DP Guns and CIWS ... You could even literally scramble helos with door guns to turkey shoot them.


Guidance -
Autonomous drones or FPV/Man-in-the-Loop drones?
If it's FPV like a lot of current drone strikes that is creating the buzz you are on, how are you going to relay guidance all the way to where the blue water 055 is? In a hostile EW environment?
If self-guided/autonomous, you've effectively made your drones into mini-AShMs = very costly and doubled/tripled/[insert ballpark figure here] the cost of your baseline AShM cruise missile. All for nothing if this drone carrier missile gets taken out before it can deploy it's load.


Lastly ... what effect on target do you hope to have with mini warheads?


I fear you are confusing the effect of drones on
land warfare - large area, spread out targets, limited GBAD assets to cover everything in that area, plenty of cover for drone operators to get close enough.
fleet defense - I am the target, optimal placement of AD assets literally at the target, very little cover for drone operators to get close enough.

bebops said:

055 is great against warship vs warship. but

In the modern warfare, they need a warship that can deal with hundreds of drones and UUV at once. Perhaps a warship with multiple laser system and CIWS.

A single warship that can destroy swarm satellite, water drones and air drones..

Click to expand...

The idea of a drone swam gets bandied around a lot, but I question the veracity of that concept. How would a swarm of drones show up in a high intensity fight hundreds of miles away from any land base (taking the Taiwan strait scenario). Any drone that can fly hundreds of miles unaided must be somewhat large. And since they will be somewhat slow, they will be found and take care way before becoming a factor. Logistic wise, if you consider a swarm means a hundred drones, say, then each one of those drone must take off somewhere, fly to the fight say 500 miles away, be controlled by somebody (somebodies) with BLOS means. And If that base is a floating one, then that becomes a glaring single point of failure. So I ,personally, am not clear how a drone swarm scenario would manifest itself.

LuzinskiJ said:

The idea of a drone swam gets bandied around a lot, but I question the veracity of that concept. How would a swarm of drones show up in a high intensity fight hundreds of miles away from any land base (taking the Taiwan strait scenario). Any drone that can fly hundreds of miles unaided must be somewhat large. And since they will be somewhat slow, they will be found and take care way before becoming a factor. Logistic wise, if you consider a swarm means a hundred drones, say, then each one of those drone must take off somewhere, fly to the fight say 500 miles away, be controlled by somebody (somebodies) with BLOS means. And If that base is a floating one, then that becomes a glaring single point of failure. So I ,personally, am not clear how a drone swarm scenario would manifest itself.

Click to expand...

Right, that's the issue. There are probably many band-aid solutions that are only effective WRT specific circumstances.

But the only real long term solution that has general application is to develop a robust Reinforcement Learning system, that is able to effectively direct the drone itself, and make critical decisions on the local system. There are still pretty large hurdles to overcome before there can be a design that reaches that level.

You would probably need a model-based RL system, since the amount of real world training episodes would be limited. By "model-based", I mean the existence of a high fidelity simulation model that takes into account realistic physics, geography, and other agents (enemy systems, aircraft, birds, etc). And your learning system must have some information gathering capacity included in a good exploration policy. Exploration policies generally should give you a ergotic distribution of relevant trajectories, which would be extremely challenging if you are talking about a space of hundreds of kts in expanse. Whatever simulator used must be capable of all different conceivable environmental dynamics, and complex reward landscape.

The optimal policy of the agent itself would need to have a very high dimensionality, in order to have the capacity to distinguish among many different sets of trajectories. Simple feedback controllers that are common in current robotics systems are far less capable than what's necessary for that size / diversity of space, and for interacting with other intelligent agents (again like enemies, birds, insect swarms, etc). So you are likely talking about large neural nets with at least 100s of billions of weight parameters. It would also need to be able to generate intrinsic stochasticity, as to not to fall for simple countermeasures. So you are talking about a large portion of these being actual Bayesian latent parameters coming from cojugate prior distributions, raising the actual parameter count by another 2-3 orders of magnitude.

Then you would probably need a full actor-critic system, so would need some good designs of value functions. And if you are talking about most modern methods, it would be an advantage function that requires both Q and state-value estimations in real time during the training process. And since you need to eventually run the agent in air-gapped mode, then you would need to perform the trajectory sampling to train the aforementioned simulator in that same mode to achieve high fidelity in the first place.

So you are basically talking about strapping a small super-computer worth of compute throughput, and a power source that can feed a small apartment building onto whatever airframe that you have designed, in order to just go through the training procedure. It's basically an insurmountable set of problems to currently be solved. Just like the promise of fully autonomous vehicles that can do all things human drive does, this too would be just as much of a dream ATM.

One of the 052DG DDG of the latest batch in Dalian sailing for her sea trails. Posted by @xx笑而不语xx on Weibo.

One of the Jiangnan units being fitted out.

by78 said:

One of the Jiangnan units being fitted out.

Click to expand...

Also noticeable is the Type 382 radar fitted on top of the mast of the currently fitting-out 052-mod cutter in the background of the 1st photo.

That it to say: Even one additional such unit being built and put to service is one additional pair of eyes-&-ears being added to complement and upgrade China's overall warfighting capability within and along the 1IC. Hence, there's no need to bemoan "why is China ordering more 054-mod and 052-mod CCG cutters instead of more proper DDGs and FFGs?" Even having only a pair of eyes and ears are still just as useful as having a pair of sword-wielding hands.

Let alone the fact where compared to proper DDGs and FFGs for the navy - Cutters for the coast guard can be built with significant portions of the ship being semi-civilian grade + without expensive weapons and associated equipment = cheaper procurement and operational costs + can be built much quicker + can be built in large numbers.

Wait, is that photoshopped in some way, or has the image processing while taking the photo with a phone gotten to such a level that it automatically does that to an image? There is like zero texture there, everything is super smooth, lines are artificially smooth, it's like I'm looking at some cell shaded CG 3D model. I understand that the zoom level is likely huge so there's gonna be a bunch of noise, but is this photo processing really the lesser of two evils there? I'd much rather have the noise and blur than this...

Totoro said:

Wait, is that photoshopped in some way, or has the image processing while taking the photo with a phone gotten to such a level that it automatically does that to an image? There is like zero texture there, everything is super smooth, lines are artificially smooth, it's like I'm looking at some cell shaded CG 3D model. I understand that the zoom level is likely huge so there's gonna be a bunch of noise, but is this photo processing really the lesser of two evils there? I'd much rather have the noise and blur than this...

Click to expand...

Real photo.

Post-processing is baked into imagery to an annoying degree nowadays plus must people then also tend to apply filters etc afterwards. This looks like it was run through a mild rotoscope filter.

She is on sea trial.